Abstract Sensory perception requires the coordinated activity of tens of thousands of neurons, working together in large- scale functional networks. Developmental events define and constrain the ultimate capabilities of these networks, therefore it is essential to understand the mechanisms underlying their formation. In the visual cortex of primates and carnivores, columnar networks of orientation preference extend across millimeters of cortical surface. Prior to eye-opening, such large-scale networks are already evident in correlated spontaneous activity, whose structure can predict future visually-evoked responses. However, a major gap in our ability to relate early network structure to mature sensory function is the lack of knowledge of the circuit mechanisms through which millimeter- scale correlated activity is generated in the early cortex. The experiments in this proposal will address this gap and test the hypothesis that large-scale correlated networks in the early cortex are generated by propagating activity through purely short-range intracortical connections and are refined through the emergence and elaboration of monosynaptic horizontal projections. By employing wide-field calcium imaging in early visual cortex, the experiments in this proposal will directly assess propagation in early spontaneous activity. A key prediction of our hypothesis is the interdependence of cortical domains in generating correlated activity. We will test this by determining whether local pharmacological or optogenetic silencing of cortical activity leads to a global disruption of correlated network activity. In order to identify the contribution of inhibitory neurons to the structure of correlated networks in the early cortex, we will utilize inhibitory-neuron specific labelling and 2-photon calcium imaging, together with targeted optogenetic manipulations. Finally, as these early correlated networks undergo refinement during the same period that long-range horizontal projections begin to elaborate in layer 2/3, we will combine anatomical labeling with longitudinal functional imaging of spontaneous activity. This will allow us to determine whether emerging horizontal connections link domains that are already functionally correlated, or instead act to reshape functional networks in the early visual cortex. Together, these studies will provide critical new insights into the circuit mechanisms governing the formation and refinement of large-scale correlated networks in the early cortex. In doing so, they will provide a key framework for understanding how the patterns of early cortical activity establish the organization of developing cortical networks and impact later sensory perception.